AI Models Show Dangerous Susceptibility to Medical Misinformation, NYU Study Finds
Research reveals AI language models can be compromised by just 0.001% of false medical data, resulting in incorrect medical advice over 7% of the time. The concerning findings show standard testing fails to detect affected systems, highlighting urgent need for better safeguards in healthcare AI.
OpenAI's Data Deletion Mishap Complicates New York Times Copyright Lawsuit
OpenAI's accidental deletion of potentially crucial evidence has added a new twist to their ongoing legal battle with major news organizations over AI training practices. The New York Times and Daily News allege unauthorized use of their content, while the data loss raises concerns about evidence preservation in technology-related disputes.
OpenAI's Data Deletion Blunder Complicates New York Times Copyright Lawsuit
OpenAI engineers accidentally erased crucial search data that could have served as evidence in an ongoing copyright lawsuit with major news publishers. The incident has forced legal teams to restart their investigation into how publishers' content was used in training AI models, highlighting challenges in examining AI training datasets.